Yesterday night, @AnthropicAI was hit with US export control restrictions around their frontier AI models. 🇺🇸
With these recent export control restrictions, Anthropic may need to verify whether a user is a US national to access certain models or capabilities.
I built a small prototype exploring how this can be done using @worldnetwork.
The flow:
1. User signs in with World ID
2. World verifies they’re a unique human
3. A passport check proves they're a US national.
4. The app receives only a yes/no proof, and not the user’s passport, name, or raw identity data
The idea is simple. AI apps should be able to enforce access rules without becoming full KYC companies or storing sensitive identity documents themselves.
Compliance will matter more as AI systems get more powerful, but the best version of compliance should be privacy-preserving by default.
"Per Axios, the government contacted Anthropic and asked to "pause releasing the... models but was unsuccessful" - i.e., Anthropic told the government to pound sand."
It's safe to assume that, if GPT-5.6 is near Mythos / Fable level, the USG has asked OpenAI to pause the release of GPT-5.6 until it can harden the national security apparatus.
Possibly the same situation for Gemini 3.5 Pro.
There were rumors that all 3 models would be released last week, yet Fable is the only one that came.
The real tin-foil hat theory is that this was Anthropic's plan. Anthropic recently stated that there could be need for a pause on AI development.
Could Anthropic telling the government to pound sand have been their gambit to force a pause?
It's a theory.
🚨 The Trump Administration tried to get Anthropic to pause the release of Fable but was unsuccessful, per a US government official
The model "needs to remain locked down" until the US security apparatus is hardened, they said, and that this "could happen in the next few weeks"
[via Axios]
Alphabet's Google will appeal a German court ruling which said it is legally liable for false claims appearing in AI Overviews, the U.S. tech company said.
https://t.co/H36BNlKVff
this is not a good sign.. the US government is making two mistakes here:
1. using "foreign national" as the criteria to gate the model is just not very smart. it's clearly not enforceable in practice and would obviously lead to a wide ban. and it's also pretty useless - there are plenty of americans more hostile towards the US than foreigners.. and it's also easy to bypass by people with real malicious intent
2. based on anthropic's post, the government clearly did not collaborate with anthropic to deeply understand the vulnerability before ordering the shutdown. this feels like a panic from people who simply don't understand what's going on
my prediction is the model access will recover early next week once the government get their sh*t together
i would not dunk on anthropic for this shutdown here. they worked hard on this release and it's a good model. they did a lot more on safety (i'd even say too much) than others. perhaps their "model is too dangerous" narrative is backfiring but still i think this shutdown is saying more about the government than anthropic
Source: Howard Lutnick told Anthropic that Mythos 5 and Fable 5 would be subject to export controls after a company claimed it was able to jailbreak Mythos (Axios)
(Visit Techmeme dot com for the link and full context!)
AI security issues must be taken seriously, but an arbitrary measure like this will put in question reliance on 🇺🇸 LLM’s and cause numerous actors to look more closely at 🇨🇳 open source solutions. https://t.co/EIjfmo26zB
The results of the US government banning Fable 5:
1. Precedent for banning AI models from ie. @OpenAI, @Google, etc
2. Countries forced to make their own AI models instead of buying from the US
3. Everyone loses, and the US surrenders its place as the world's undisputed economic superpower
The US and the US dollar have ruled the world because of its technology EXPORT (!!)
The world loved to buy and run on American tech!
Now, this trust is eroding
It is now up to companies like Strawberry and others to build our own AI models and platforms
we've never been more ready
It might take some time, but it's coming
And when it's here, it will be glorious
OpenAI is under investigation by a coalition of state attorneys general who requested information from the artificial intelligence company on a wide range of topics. https://t.co/dBtOM29hIc
COMMENTARY: When news broke that the Pentagon wouldn’t sell long-range Tomahawk missiles to Germany, it suggested Washington might fear Moscow would view such a capability in Europe’s hands as dangerous escalation.
https://t.co/h48CKiMfOJ
The US has started banning Europeans from using their latest AI models.
The EU needs sovereign models that compete, otherwise we will be cut off.
Mistral is perfectly positioned to deliver, European governments must choose it, and help it thrive with less burdensome rules.